Tennis digital teaching resource management system under smart sports environment

Abstract

The application discloses a tennis digital teaching resource management system in a smart sports environment and relates to the technical fields of smart sports and digital teaching. The system synchronously collects visual data, inertial measurement data, racket sensing data and ball movement data of students in the process of tennis training or competition through a data acquisition module, and uses a multi-modal data fusion analysis engine to perform unified space-time alignment and fusion processing on the data to generate unified scene data packets corresponding to single ball hitting events. The system generates semantic labels based on the unified scene data packets and realizes intelligent matching and personalized generation of teaching resources accordingly. Related data is stored in a structured resource database and training analysis and feedback are provided to coaches and students through an application service terminal, so that the intelligentialization and pertinence of tennis teaching are improved.

Description

Technical Field

[0001] This invention relates to the field of smart sports and digital teaching technology, and more specifically, to a digital teaching resource management system for tennis in a smart sports environment. Background Technology

[0002] Tennis, as a competitive sport that demands high precision in technique, body coordination, and tactical awareness, typically relies on the coach's on-site observation, experience-based judgment, and the student's subjective feelings during its teaching and training. In traditional tennis instruction, coaches primarily observe students' hitting motions, shot effectiveness, and rally performance, providing verbal explanations or demonstrations to correct technical problems. While this teaching method has accumulated a wealth of experience over time, it is highly dependent on the coach's individual ability, making it difficult to guarantee the objectivity and consistency of teaching conclusions. This is especially true in scenarios with a large number of students or a fast-paced training environment, where incomplete analysis and untimely feedback are prone to occur.

[0003] With the development of video capture and wearable sensing technologies, some teaching systems have attempted to incorporate video playback, motion capture, or single-sensor data to assist tennis training. For example, single-camera video playback helps students observe their own movements, or racket sensors track parameters such as the number of hits and the force of the shot. However, most of these solutions focus on a single data source and lack a unified model of the overall relationship between human movement, racket behavior, and ball motion, making it difficult to accurately reflect the synergistic relationship between various elements during a shot. In practical applications, these systems often only provide fragmented data indicators or static playback, failing to offer clear and interpretable analytical results for complex technical problems (such as insufficient body rotation, incorrect shot timing, and broken motion chains).

[0004] On the other hand, the organization and use of existing tennis teaching resources are rather rudimentary. Instructional videos, technical explanations, and training plans typically exist as fixed courses or manually categorized, lacking dynamic relevance to the specific training scenarios and technical problems of students. When a student makes a particular mistake during training, coaches or students often need to manually sift through a large amount of teaching resources, which is not only inefficient but also makes it difficult to ensure that the selected resources are highly relevant to the current training problem. Furthermore, traditional teaching methods struggle to systematically accumulate and structure student training data over a long period, making it difficult to distill training experience into reusable teaching knowledge and limiting the possibility of optimizing teaching and conducting tactical analysis based on historical data.

[0005] Therefore, how to achieve unified collection and fusion analysis of multi-source training data during tennis training, how to transform scattered motion data and ball movement data into understandable and comparable training scenario information, and how to achieve intelligent generation and precise delivery of teaching resources based on training scenarios and technical issues have become urgent technical problems to be solved in the field of digital tennis teaching.

[0006] Therefore, there is an urgent need for a digital teaching resource management system for tennis in a smart sports environment to solve these problems. Summary of the Invention

[0007] The purpose of this invention is to solve the technical problems mentioned in the background section and to provide a digital teaching resource management system for tennis in a smart sports environment.

[0008] The objective of this invention is achieved through the following technical solution: A digital teaching resource management system for tennis in a smart sports environment includes: The data acquisition module is used to synchronously collect multimodal training data of trainees during tennis training or matches, with the ball-hitting event as the basic sampling unit. The multimodal training data includes at least visual data, inertial measurement data, racket sensing data, and environmental perception data used to describe the ball's motion state. The multimodal data fusion and analysis engine is connected to the data acquisition module and is used to perform time alignment, spatial calibration and fusion processing on the multimodal training data based on a unified spatiotemporal reference, construct a three-dimensional human motion model corresponding to a single ball hitting event, and associate the three-dimensional human motion model with the ball motion information to generate a unified scenario data package representing the training process. The AI-based contextualized resource generation module is communicatively connected to the multimodal data fusion and analysis engine. It is used to intelligently analyze the unified contextual data package, generate semantic tags describing technical actions, action quality, error types, and training scenarios, and retrieve, combine, or dynamically generate teaching resources based on the semantic tags. A structured resource database is used to store standard teaching resources, historical training scenario data packages, and personalized teaching resources generated by the AI ​​contextualized resource generation module, and establishes an association index relationship between resources through the semantic tags; The application service terminal is used to provide coaches and trainees with visual displays of the training process, feedback on analysis results, and interactive and push functions for personalized teaching resources.

[0009] As a preferred embodiment of the present invention, the data acquisition module includes: Multi-view camera equipment deployed at the training ground is used to collect visual data on trainees and the movement of the ball; Inertial measurement units are installed on key parts of the student's body and on the racket to collect data on the student's body posture, acceleration, and angular velocity. Sensors integrated into the racket are used to collect vibration characteristics, racket face angle, or swing path information at the moment of impact. And radar sensors or high-speed optical sensors used to collect information on ball speed, trajectory, or landing point.

[0010] As a preferred technical solution of the present invention, the multimodal data fusion analysis engine is used for: Establish a unified timeline based on the moment of impact to synchronize the time of multimodal training data from different sensor sources; A unified spatial coordinate system is constructed to spatially calibrate the visual data, inertial measurement data, and racket sensing data; Based on the fused data-driven human biomechanical model, a three-dimensional skeletal motion model of the trainee during the ball-hitting process is reconstructed, and the ball's trajectory is associated with the three-dimensional skeletal motion model to form the unified scenario data package.

[0011] As a preferred technical solution of the present invention, the AI ​​contextualized resource generation module includes: The intelligent tagging unit is used to analyze the unified scenario data package based on a deep learning model or a rule model and automatically generate at least one semantic tag, which includes a technical action category tag, an action quality assessment tag, an error type tag, or a training scenario tag. The contextualized slicing and comparison unit is used to divide the continuous training process into multiple independent training scenario segments according to the semantic tags or preset rules, and to compare and analyze the trainee's training scenario segments with standard teaching resources or historical excellent examples.

[0012] As a preferred technical solution of the present invention, the contextualized slicing and comparison unit is also used to generate an augmented reality comparison view. The augmented reality comparison view displays the standard action trajectory, force direction or ideal hitting position in the corresponding training scenario video of the trainee in a superimposed manner, so as to intuitively show the difference between the trainee's action and the standard action.

[0013] As a preferred technical solution of the present invention, the AI ​​contextualized resource generation module further includes: The personalized content generation unit is used to automatically retrieve and combine multiple basic teaching resources from the structured resource database based on the high-frequency error types or technical shortcomings identified in the semantic tags, and generate targeted personalized error correction training programs; and / or generate virtual opponent models with specific tactical behavioral characteristics based on the analysis of trainees' tactical execution patterns, for use in tactical confrontation training in a simulated or simulated environment.

[0014] As a preferred embodiment of the present invention, the structured resource database is a database built based on a knowledge graph, wherein: Each teaching resource, training scenario data package, or error type is treated as a knowledge node. The nodes are connected through the associations defined by the semantic tags to form a knowledge network that describes the relationships between technologies, the learning progression relationships, and the error correction logic relationships.

[0015] As a preferred technical solution of the present invention, the application service terminal includes: The coach's command platform is used to display multimodal data, 3D motion models, key technical indicators, and system-generated teaching suggestions during the training process. The student interaction terminal is used to present the training analysis results, the augmented reality comparison view, and the personalized error correction training scheme to the student.

[0016] This invention also provides a method for managing digital tennis teaching resources, comprising the following steps: S1: Through the data acquisition module, multimodal training data of trainees during training or competition is collected synchronously; S2: The multimodal training data is fused and processed by the multimodal data fusion analysis engine to generate a unified scenario data package containing three-dimensional action reconstruction; S3: The unified contextual data package is analyzed and multi-dimensional semantic tags are generated through the AI ​​contextual resource generation module; S4: Based on the semantic tags and students' historical training data, match or dynamically generate personalized teaching resources from the structured resource database; S5: The unified context data package, semantic tags, and personalized teaching resources are associated and stored, and then pushed to the corresponding application service terminal.

[0017] As a preferred technical solution of the present invention, the dynamic generation of personalized teaching resources in step S4 includes: Based on error type tags, generate electronic lesson plans containing targeted practice content, training points, and training plans; and / or based on tactical analysis results, drive the virtual opponent model to execute preset or adaptive tactical sequences in a simulation environment to conduct interactive tactical drills with trainees.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a multimodal data fusion mechanism centered on the ball-hitting event. It unifies and correlates visual data, inertial measurement data, racket sensor data, and ball motion data generated by trainees during tennis training or matches, forming a unified contextual data package capable of simultaneously representing human movements and ball motion states. Compared to traditional teaching methods that rely solely on single videos or experience-based judgments, this invention achieves an objective quantitative description of the entire technical movement process, significantly improving the accuracy and stability of movement analysis. This ensures that teaching evaluation results no longer depend on the coach's personal experience, providing a repeatable, traceable, and comparable technical foundation.

[0019] This invention utilizes an AI-powered contextualized resource generation module to intelligently analyze unified contextualized data packages, automatically generating semantic tags describing action categories, action quality, error types, and training scenarios. These semantic tags serve as the core index for organizing and retrieving teaching resources, enabling a shift in teaching resource allocation from "content-based classification" to "precise matching based on technical issues." The system can automatically retrieve, combine, and generate personalized error-correction training plans tailored to specific technical weaknesses of learners. Furthermore, through contextualized slices and augmented reality comparison views, learners can intuitively compare their own actions with standard examples, significantly improving their understanding of error causes and improvement directions, and shortening the technical correction cycle.

[0020] This invention employs a structured resource database based on knowledge graphs to systematically model the relationships between training scenario data, technical actions, error types, and teaching resources. This ensures that each training result is not only stored but also continuously utilized within an evolving knowledge network. Through this structured management approach, the system supports long-term training data accumulation, teaching pattern mining, and group ability analysis. Furthermore, it supports advanced applications such as virtual opponent generation and tactical simulation training, comprehensively enhancing the systematic, forward-looking, and intelligent level of tennis instruction, and driving the development of tennis teaching from experience-driven to data-driven and intelligently assisted approaches. Attached Figure Description

[0021] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with embodiments and appendices. Figures 1-2 The present invention will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0023] Example 1: In this example, the system is deployed on a standard tennis training court and includes a data acquisition module, a multimodal data fusion and analysis engine, an AI-based contextualized resource generation module, a structured resource database, and an application service terminal. The system uses a single shot event as the basic unit for training analysis, and comprehensively models and manages the trainee's movements, ball motion, and training scenarios during the training process.

[0024] Before training begins, the coach sets the training mode through the coach command console in the application service terminal, such as forehand multi-ball training, baseline rally training, or serve-and-volley training, and creates corresponding training tasks for the trainees. After the system starts, the data acquisition module is activated as a whole, synchronously collecting multimodal data of the trainees during the training process. The data acquisition module includes multi-view camera equipment deployed at different locations on the training field to collect visual data of trainee movements and ball motion; inertial measurement units worn on key parts of the trainee's body and racket to collect trainee body posture and racket swing information; sensors integrated on the racket to collect vibration characteristics, racket face angle, and hitting position at the moment of impact; and radar or high-speed optical sensors set around the court to collect ball speed, ball trajectory, and landing point information. In this invention, the inertial measurement unit is a wearable inertial sensing module, which includes at least a three-axis accelerometer and a three-axis gyroscope, and preferably also includes a three-axis magnetometer, thus forming a nine-axis inertial measurement unit (IMU). The inertial measurement unit can collect information on the spatial attitude change, angular velocity change, and linear acceleration change of the wearing part in real time during training, and send the collected data to the multimodal data fusion analysis engine through wireless communication.

[0025] In practical implementation, the inertial measurement unit can be implemented using existing mature miniature inertial sensor modules, such as an IMU module integrated into a flexible wearable shell. The shell is fixed to the key movement parts of the student's body and the racket by straps, adhesive structures, or embedding methods. The key body parts include at least one or more of the torso, upper arm, forearm, and wrist, to reflect the student's body rotation, swing trajectory, and power sequence during the hitting process.

[0026] At the racket end, the inertial measurement unit is preferably installed inside the racket handle or at the end of the handle, forming an integrated structure with the racket through embedded installation or modular installation. It is used to collect the changes in attitude, angular velocity and acceleration of the racket during the swing and hitting process, thereby characterizing the racket's swing speed, swing plane and dynamic characteristics at the moment of impact.

[0027] The inertial measurement unit (IMU) incorporates a data acquisition circuit, a clock synchronization module, and a wireless communication module. Its sampling frequency can be set according to training needs to meet the requirements of high-speed motion acquisition during tennis shots. By synchronizing the data acquired by the body IMU and the racket IMU, the system can accurately reflect the temporal and spatial relationships between human movements and racket motion.

[0028] The multimodal data fusion and analysis engine communicates with the data acquisition module, receiving various sensor data in real time during training. Using the moment of impact as a unified reference, the engine performs time synchronization and spatial calibration on data from different sources, and constructs a 3D human motion model of the trainee during each shot based on the fused data. Simultaneously, the multimodal data fusion and analysis engine correlates the trainee's motion model with the ball's trajectory and racket impact information, generating a unified scenario data package corresponding to a single shot event, used to fully describe the technical movements and training scenario of that shot.

[0029] The generated unified scenario data package is transmitted to the AI-based contextualized resource generation module for further processing. The intelligent tagging unit within the AI-based contextualized resource generation module analyzes the unified scenario data package and automatically generates semantic tags describing the type of technical action, the level of action quality, the type of error, and the training scenario. These semantic tags are used to characterize the trainee's technical performance and any problems encountered during that shot.

[0030] Building upon this foundation, the contextualized slicing and comparison unit in the AI-based contextualized resource generation module divides the continuous training process into multiple independent training scenario segments based on semantic tags. It then compares and analyzes these training scenario segments with standard teaching examples or outstanding student examples stored in the structured resource database. The comparison results can be presented using augmented reality, such as overlaying standard movement trajectories, ideal force application paths, or recommended hitting points onto the student's own training video, creating an augmented reality comparison view.

[0031] Simultaneously, the personalized content generation unit in the AI-based contextualized resource generation module automatically retrieves and combines multiple basic teaching resources from the structured resource database based on the high-frequency error types or technical shortcomings reflected in the semantic tags, generating a personalized error correction training plan for the student. This error correction training plan may include technical point explanation videos, specific physical or technical practice content, and on-field training configuration suggestions, presented in the form of a structured electronic lesson plan.

[0032] The structured resource database is used to centrally store standard teaching resources, historical training scenario data packages, and personalized teaching resources generated by the system. This database adopts a knowledge graph-based organization method, using technical actions, error types, training scenarios, and teaching resources as knowledge nodes. Semantic tags define the relationships between nodes, describing technical connections, learning progression relationships, and error correction logic relationships.

[0033] The application service terminal provides system interaction functions for coaches and trainees. The coach's console can display multiple video feeds, 3D motion models, key training indicators, and system-generated teaching suggestions in real time during the training process; the trainee interaction terminal is used to display individual training analysis results, augmented reality comparison views, and personalized error correction training plans to trainees, thereby achieving instant feedback and teaching guidance during the training process.

[0034] As can be seen from the above system embodiments, the system of the present invention can unify and integrate scattered training data and manage it in a structured manner, realizing the transformation of tennis teaching from experience-driven to data-driven and intelligently assisted.

[0035] Example 2: In this example, the method is applied to daily tennis training courses to analyze the technical movements and tactical execution of students and generate corresponding teaching resources.

[0036] At the beginning of the training phase, multimodal training data is synchronously collected from trainees through a data acquisition module. This includes visual data of trainee movements, inertial measurement data of the body and racket, racket hitting characteristic data, and ball trajectory data. The multimodal data is synchronized in time during acquisition to ensure consistency between different data sources.

[0037] Subsequently, the collected multimodal training data is fused and processed using a multimodal data fusion analysis engine. During the fusion process, the system uses a single shot event as the unit of analysis, and performs correlation modeling on the student's action data and ball motion data before and after the shot, generating a unified scenario data package that can simultaneously describe the student's technical actions and the ball's motion state.

[0038] Next, the unified scenario data package is analyzed and processed through the AI-based contextualized resource generation module. The system identifies the trainee's technical actions and evaluates the quality of action completion and potential errors. At the same time, it generates multi-dimensional semantic tags that include action category, error type, quality evaluation, and training scenario, based on the current training task and training mode.

[0039] Based on this, the system intelligently matches relevant teaching resources from a structured resource database according to the generated semantic tags and the trainees' historical training data, or dynamically generates personalized teaching resource content according to current training needs. These personalized teaching resources may include targeted practice videos, training point explanations, training plans, and tactical exercise suggestions.

[0040] In some training scenarios, the system can also generate virtual opponent models with specific tactical behavioral characteristics based on the analysis of trainees' tactical execution patterns. In a simulated or analog environment, the virtual opponent can engage in adversarial training with trainees according to preset or adaptive tactical strategies, thereby helping trainees to strengthen their practice against specific tactical weaknesses.

[0041] Finally, the system associates and stores the generated unified scenario data package, semantic tags, and personalized teaching resources, and pushes them to the corresponding application service terminals. Coaches can view the training analysis results and make teaching adjustments through the coach command console, while trainees can view their personal training feedback and training plans through the trainee interaction terminal, thus completing a full training analysis and teaching feedback process.

[0042] As can be seen from the above method embodiments, the method of the present invention can realize the automatic collection, intelligent analysis and dynamic generation of tennis training data and teaching resources, forming a closed loop of digital teaching management that runs through the pre-training, training and post-training stages.

[0043] Example 3: In this embodiment, the system takes a single ball-hitting event as the smallest unit of analysis. Through the collaborative work of the data acquisition module, the multimodal data fusion analysis engine, the AI ​​contextualized resource generation module, the structured resource database, and the application service terminal, the system achieves closed-loop operation of training data acquisition, motion analysis, teaching resource generation and management.

[0044] Before training begins, the coach sets the training mode, technical movement type, and training objectives through the coach command console in the application service terminal. After the system starts, the data acquisition module is activated to synchronously collect multimodal training data from trainees during training or competition. The data acquisition module includes multi-view camera equipment deployed on the training field, inertial measurement units worn on key parts of the trainees' bodies and rackets, sensors integrated into the rackets, and radar or high-speed optical sensors for collecting the ball's motion. All acquisition devices are synchronized through a unified network to ensure that data from different modalities are processed under the same time reference.

[0045] Let the time of contact of the ball in the k-th stroke be . The system selects a time window of length Δt before and after the shot as the analysis interval, i.e.:

[0046] Where t represents the sampling time, and Δt is in seconds. Within this time window, the visual data acquired by the multi-view camera device is represented as follows:

[0047] in, Let be the image frame captured by the c-th camera at time t, where C is the number of cameras. The body and racket motion data output by the inertial measurement unit are represented as follows:

[0048] in, Let be the attitude angle vector of the j-th joint or sensor node. It is the angular velocity vector. Let J be the linear acceleration vector, and J be the number of inertial sensor nodes. The hit characteristic data collected by the racket sensor is represented as:

[0049] in, These are the characteristic parameters of the impact vibration. For the angle parameter of the racket face, Let be the position vector of the hitting point in the racket face coordinate system.

[0050] The motion of the sphere is acquired by radar or a high-speed optical sensor, and its trajectory is represented as follows:

[0051] in, These represent the spatial position components of the sphere in a unified world coordinate system. The multimodal data fusion analysis engine uses the moment of impact as its starting point. As a unified time reference, the aforementioned multimodal training data undergoes temporal alignment and spatial fusion processing. First, temporal normalization is performed:

[0052] Map all data to a unified relative time axis .

[0053] Subsequently, pose estimation was performed on the visual data to obtain the coordinates of the two-dimensional joints:

[0054] in, Let be the two-dimensional coordinates of the j-th joint from the c-th viewpoint. This is a pose estimation model. Through camera calibration and multi-view reconstruction, 2D keypoints are mapped to visually estimated 3D keypoints.

[0055] Simultaneously, the inertial estimation of the joint states is constructed based on the inertial measurement data:

[0056] The system employs a sensor fusion model to combine visual and inertial estimations to obtain the final 3D joint position:

[0057] in, and Let be the weight matrix, satisfying This allows us to construct a three-dimensional skeletal motion model of the student during the k-th shot:

[0058] The multimodal data fusion analysis engine will integrate the three-dimensional skeletal motion model. Sphere trajectory and racket sensor data The data is associated and encapsulated to form a unified contextual data package corresponding to a single ball-hitting event:

[0059] The unified contextual data package is transmitted to the A contextualized resource generation module for analysis and processing. The intelligent tagging unit in the A contextualized resource generation module classifies the technical actions based on the three-dimensional action sequence.

[0060] in, For action category labels.

[0061] Meanwhile, the system calculates key technical parameters based on a biomechanical model, such as the trunk rotation angle:

[0062] And extract the peak value of the rotation angle:

[0063] The peak rotation angle is compared with the standard rotation angle of the corresponding action in the standard model. By comparing the results, we can obtain the relative deviation:

[0064] when At that time, the system determined that there was an insufficient rotation problem and generated a corresponding error type label. ,in Combined action category labels with preset threshold parameters Error type label Action quality evaluation results and training mode labels This forms the semantic tag set for the ball-hitting event:

[0065] in Action category label (e.g., "forehand drive / serve / volley"), discrete symbol; Error type label (e.g., "insufficient rotation / too late hitting point"), discrete symbol; Action quality level label (such as "Excellent / Good / Medium / Poor" or quantitative level), discrete or ordered discrete quantity; Training scenario labels (e.g., "multi-ball fixed-point shooting / competitive round / tactical drill"), discrete symbols. It is a set / vector label container, dimensionless.

[0066] The contextualized slicing and comparison unit in the AI ​​contextualized resource generation module slices the continuous training data according to the semantic tag set, aligns and compares the training scenario segments corresponding to the trainees with the standard example segments stored in the structured resource database, and generates an augmented reality comparison view, which presents the standard action trajectory or ideal force path in the trainee's training video in an overlay manner.

[0067] Meanwhile, the personalized content generation unit in the AI ​​contextualized resource generation module generates content based on the semantic tag set. It retrieves basic teaching resources from a structured resource database that match error types and technical weaknesses, and generates personalized error correction training schemes through combinatorial functions:

[0068] in, It is a collection of teaching resources in the database. ,in Let be the i-th resource unit (micro-lesson video, standard example clip, special exercise, training plan template, AR overlay template, tactical library item, etc.), and N be the number of resource units; For a set of objects, dimensionless. The personalized teaching resource output package (also known as the "personalized error correction training plan") corresponding to the k-th hit event may include electronic lesson plan items, training points, recommended practice video links, on-court training plan parameters, AR comparison view generation instructions, etc. A structured object / collection (non-scalar), dimensionless. k: index of the ball-hitting event, k=1,2,…

[0069] Personalized content generation function / generation strategy operator, used to complete resource retrieval, filtering, combination and arrangement output based on tags and resource library information; G is the mapping relationship (which can be implemented by rule engine, learning model or a combination of the two), and is dimensionless.

[0070] The generated personalized error correction training plan, augmented reality comparison view, and corresponding analysis results are pushed to the coach's command console and the student's interactive terminal via the application service terminal, enabling training feedback and teaching guidance. After training, a unified scenario data package is generated. Semantic tag set All generated teaching resources are stored in a structured resource database. This database employs a knowledge graph-based organization, using training scenarios, technical actions, error types, and teaching resources as knowledge nodes. It constructs a technical association network, a progressive relationship network, and an error correction logic network through semantic tag-defined relationships to support subsequent training analysis and teaching decisions. As can be seen from the above implementation, this invention, through the collaborative work of the various module units defined in the claims, achieves multimodal data fusion analysis and intelligent generation of teaching resources centered on the ball-hitting event, thereby constructing a complete process for tennis teaching data collection, analysis, feedback, and resource management.

[0071] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but the present invention is not limited to these embodiments. Equivalent modifications made by those skilled in the art without departing from the principles of the present invention should fall within the protection scope of the present invention.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital teaching resource management system for tennis in a smart sports environment, characterized in that, include: The data acquisition module is used to synchronously collect multimodal training data of trainees during tennis training or matches, with the ball-hitting event as the basic sampling unit. The multimodal training data includes at least visual data, inertial measurement data, racket sensing data, and environmental perception data used to describe the ball's motion state. The multimodal data fusion and analysis engine is connected to the data acquisition module and is used to perform time alignment, spatial calibration and fusion processing on the multimodal training data based on a unified spatiotemporal reference, construct a three-dimensional human motion model corresponding to a single ball hitting event, and associate the three-dimensional human motion model with the ball motion information to generate a unified scenario data package representing the training process. The AI-based contextualized resource generation module is communicatively connected to the multimodal data fusion and analysis engine. It is used to intelligently analyze the unified contextual data package, generate semantic tags describing technical actions, action quality, error types, and training scenarios, and retrieve, combine, or dynamically generate teaching resources based on the semantic tags. A structured resource database is used to store standard teaching resources, historical training scenario data packages, and personalized teaching resources generated by the AI ​​contextualized resource generation module, and establishes an association index relationship between resources through the semantic tags; The application service terminal is used to provide coaches and trainees with visual displays of the training process, feedback on analysis results, and interactive and push functions for personalized teaching resources.

2. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The data acquisition module includes: Multi-view camera equipment deployed at the training ground is used to collect visual data on trainees and the movement of the ball; Inertial measurement units are installed on key parts of the student's body and on the racket to collect data on the student's body posture, acceleration, and angular velocity. Sensors integrated into the racket are used to collect vibration characteristics, racket face angle, or swing path information at the moment of impact. And radar sensors or high-speed optical sensors used to collect information on ball speed, trajectory, or landing point.

3. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The multimodal data fusion and analysis engine is used for: Establish a unified timeline based on the moment of impact to synchronize the time of multimodal training data from different sensor sources; A unified spatial coordinate system is constructed to spatially calibrate the visual data, inertial measurement data, and racket sensing data; Based on the fused data-driven human biomechanical model, a three-dimensional skeletal motion model of the trainee during the ball-hitting process is reconstructed, and the ball's trajectory is associated with the three-dimensional skeletal motion model to form the unified scenario data package.

4. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The AI-based contextualized resource generation module includes: The intelligent tagging unit is used to analyze the unified scenario data package based on a deep learning model or a rule model and automatically generate at least one semantic tag, which includes a technical action category tag, an action quality assessment tag, an error type tag, or a training scenario tag. The contextualized slicing and comparison unit is used to divide the continuous training process into multiple independent training scenario segments according to the semantic tags or preset rules, and to compare and analyze the trainee's training scenario segments with standard teaching resources or historical excellent examples.

5. The tennis digital teaching resource management system in a smart sports environment according to claim 4, characterized in that, The contextualized slice and comparison unit is also used to generate an augmented reality comparison view. The augmented reality comparison view displays the standard movement trajectory, force direction, or ideal hitting position in the corresponding training scenario video of the trainee in an overlay manner, so as to intuitively show the difference between the trainee's movement and the standard movement.

6. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The AI-based contextualized resource generation module also includes: The personalized content generation unit is used to automatically retrieve and combine multiple basic teaching resources from the structured resource database based on the high-frequency error types or technical shortcomings identified in the semantic tags, and generate targeted personalized error correction training programs; and / or generate virtual opponent models with specific tactical behavioral characteristics based on the analysis of trainees' tactical execution patterns, for use in tactical confrontation training in a simulated or simulated environment.

7. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The structured resource database is a database built on a knowledge graph, wherein: Each teaching resource, training scenario data package, or error type is treated as a knowledge node. The nodes are connected through the associations defined by the semantic tags to form a knowledge network that describes the relationships between technologies, the learning progression relationships, and the error correction logic relationships.

8. The tennis digital teaching resource management system in a smart sports environment according to claim 1, characterized in that, The application service terminal includes: The coach's command platform is used to display multimodal data, 3D motion models, key technical indicators, and system-generated teaching suggestions during the training process. The student interaction terminal is used to present the training analysis results, the augmented reality comparison view, and the personalized error correction training scheme to the student.

9. A method for managing digital tennis teaching resources based on the system described in any one of claims 1-8, characterized in that, Includes the following steps: S1: Through the data acquisition module, multimodal training data of trainees during training or competition is collected synchronously; S2: The multimodal training data is fused and processed by the multimodal data fusion analysis engine to generate a unified scenario data package containing three-dimensional action reconstruction; S3: The unified contextual data package is analyzed and multi-dimensional semantic tags are generated through the AI ​​contextual resource generation module; S4: Based on the semantic tags and students' historical training data, match or dynamically generate personalized teaching resources from the structured resource database; S5: The unified context data package, semantic tags, and personalized teaching resources are associated and stored, and then pushed to the corresponding application service terminal.

10. The method according to claim 9, characterized in that, The dynamic generation of personalized teaching resources mentioned in step S4 includes: Based on error type tags, generate electronic lesson plans containing targeted practice content, training points, and training plans; and / or based on tactical analysis results, drive the virtual opponent model to execute preset or adaptive tactical sequences in a simulation environment to conduct interactive tactical drills with trainees.

Images